Study on Fruit Recognition Methods Based on Compressed Sensing

Using multiple features of the image, the recognition rate of the fruit images can be improved. But with the increase of the number of the fruits, the recognition complex also increased. For reducing the complex degree of the algorithms and meeting demand of lots of fruits recognition, a method based on compressed sensing was proposed. Sixty-three fruit images were investigated, and eight texture feature values and seven shape feature values were extracted to construct the training eigenmatrix. Based on compressed sensing, the sparse coefficient vector which was the sparse representation of the sample eigenvector on the training eigenmatrix can be obtained, so the test sample was classified by analyzing the coefficient vector. Using the different fruits upon a blank background in this paper, the recognition rates of this method are 83%. The experimental results showed that the recognition method based on compressed sensing could effectively recognize the different class fruits.